EMBO. Ligand-regulated oligomerization of b 2 -adrenoceptors in a model lipid bilayer EMBO. open

Transcription

1 (9), & 9 European Molecular Biology Organization Some Rights Reserved 6-89/9 Ligand-regulated oligomerization of b -adrenoceptors in a model lipid bilayer THE EMBO JOURNAL EMBO open This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits distribution, and reproduction in any medium, provided the original author and source are credited. This license does not permit commercial exploitation without specific permission. Juan José Fung, Xavier Deupi, Leonardo Pardo, Xiao Jie Yao, Gisselle A Velez-Ruiz, Brian T DeVree, Roger K Sunahara and Brian K Kobilka, * Department of Molecular and Cellular Physiology, Stanford University School of Medicine, Stanford, CA, USA, Laboratori de Medicina Computacional, Unitat de Bioestadística, Facultat de Medicina, Universitat Autònoma de Barcelona, Barcelona, Spain and Department of Pharmacology, University of Michigan Medical School, Ann Arbor, MI, USA The b -adrenoceptor (b AR) was one of the first Family A G protein-coupled receptors (GPCRs) shown to form oligomers in cellular membranes, yet we still know little about the number and arrangement of protomers in oligomers, the influence of ligands on the organization or stability of oligomers, or the requirement for other proteins to promote oligomerization. We used fluorescence resonance energy transfer (FRET) to characterize the oligomerization of purified b AR site-specifically labelled at three different positions with fluorophores and reconstituted into a model lipid bilayer. Our results suggest that the b AR is predominantly tetrameric following reconstitution into phospholipid vesicles. Agonists and antagonists have little effect on the relative orientation of protomers in oligomeric complexes. In contrast, binding of inverse agonists leads to significant increases in FRET efficiencies for most labelling pairs, suggesting that this class of ligand promotes tighter packing of protomers and/or the formation of more complex oligomers by reducing conformational fluctuations in individual protomers. The results provide new structural insights into b AR oligomerization and suggest a possible mechanism for the functional effects of inverse agonists. advance online publication, 7 September 9; doi:.8/emboj.9.67 Subject Categories: signal transduction Keywords: b -Adrenoceptor; FRET; inverse agonist; oligomers; TM6 Introduction The majority of physiological responses to hormones and neurotransmitters are mediated by G protein-coupled receptors (GPCRs). These integral membrane proteins relay signals *Corresponding author. Department of Molecular and Cellular Physiology, Stanford University School of Medicine, Room B57, Beckman Center, 79 Campus Drive, Stanford, CA 95, USA. Tel.: þ ; Fax: þ ; kobilka@stanford.edu Received: 6 April 9; accepted: 8 August 9 elicited by an array of structurally diverse agonists including photons, ions, small organic compounds, peptides, proteins and lipids (Hill, 6). GPCRs have been classically described as monomeric transmembrane receptors that form a ternary complex: a ligand, the GPCR, and its associated G protein (Chabre and le Maire, 5; Fung et al, 98). This classical view is compatible with observations that monomeric rhodopsin and b -adrenoceptor (b AR) are capable of activating G proteins (Bayburt et al, 7;Ernstet al, 7;Whortonet al, 7). Nevertheless, it is now well accepted that Family C GPCRs are constitutive dimers (Jones, et al, 998; Margeta-Mitrovic et al, ; Galvez et al, ),andmanyfamilyagpcrshavebeen observed to oligomerize in cells. This was initially shown for the b AR in 996 (Hebert et al, 996),andwaslaterfollowedby publications demonstrating both homo- and hetero-oligomerization of a broad spectrum of Family A GPCRs using a variety of techniques (Jordan and Devi, 999; Angers et al, ; Gines et al, ; Rocheville et al, ; Cheng and Miller, ; Mellado et al, ;Latifet al, ;Salahpouret al, ; Gonzalez-Maeso et al, 8;Guo et al, 8;Vilardaga et al, 8). Despite this evidence, the effect of ligands on formation, organization and stability of receptor oligomers, as well as the role of other cellular proteins, is not well understood and may be receptor subtype specific (Angers et al, ;Cheng and Miller, ; Latif et al, ;Zhuet al, ;RoessandSmith,; Law et al, 5). High-resolution crystal structures are now available for the inactive state of the b AR (Cherezov et al, 7; Rasmussen et al, 7; Rosenbaum et al, 7); yet, we still know very little about the structure, stoichiometry and dynamics of oligomers in lipid bilayers. To date, much of what is known about oligomerization of b ARs and other Family A GPCRs comes from elegant studies using intermolecular fluorescence/bioluminescence resonance energy transfer (FRET or BRET) in live cells (Angers et al, ; Salahpour et al, ; Milligan and Bouvier, 5). In an effort to complement these cell-based studies and better understand the process of GPCR oligomerization and the organization of protomers within oligomers, we investigated oligomerization of purified b AR in a model membrane system. By site specifically labelling purified monomeric b AR and reconstituting it into lipid vesicles, we are able to show that this receptor effectively forms specific multimeric assemblies in lipid bilayers as monitored by FRET and cross-linking studies. FRET saturation studies are most consistent with the formation of tetramers, and differences in FRET between different labelling pairs allow us to propose a model of the orientation of the protomers within the tetramer. An agonist and neutral antagonist have little effect on b AR oligomerization, but the inverse agonist ICI 8,55 promotes rearrangement of the protomers and/or the formation of higher-order oligomers. & 9 European Molecular Biology Organization

2 b -Adrenoceptor oligomers Results Site-specific labelling of purified monomeric b AR and reconstitution into lipid vesicles The goal of our studies was to monitor self-association of b AR following reconstitution in lipid vesicles and to obtain information about the relative orientation of protomers in oligomers. FRET using small-molecular-weight fluorescent probes is an ideal tool for these studies because it provides relative distance information yet requires relatively small amounts of protein (Mansoor et al, 6). To achieve sitespecific labelling of the b AR, we generated modified receptors having single-reactive cysteines that can be chemically modified with sulphydryl-reactive fluorophores. Mutants were made on a minimal cysteine background in which the five chemically reactive cysteines, out of, were mutated (see section Materials and methods). These mutations have no effect on ligand binding or G protein coupling. The remaining three cysteines that are not palmitoylated or are part of disulphide bonds are not reactive due to their location in the hydrophobic core (Figure A). We initially constructed 8 single-cysteine mutants in the cytoplasmic domains of the receptor. Three were chosen based on their functional properties, chemical reactivity and their distribution (Figure B): D5-T66C (intracellular loop-, ICL), D5-A65C (transmembrane domain-6, TM6) and D5-RC (helix-8, H8) (Cherezov et al, 7; Rasmussen et al, 7; Rosenbaum et al, 7). This spatial distribution of the labelling sites was designed to provide information about the orientation of protomers relative to each other. Modified receptors were expressed in Sf9 cells using recombinant baculovirus and purified using sequential antibody and alprenolol affinity chromatography. We have shown previously that this purification protocol produces monomeric, detergent-solubilized b AR (Whorton et al, 7). Purified, detergent-solubilized b AR was labelled with Cy or Cy5 maleimide. These fluorophores were chosen for FRET studies because they possess an R value (Förster critical distance where 5% of energy transfer occurs) in the range of 7 to 56 Å depending on the experimental system (Mansoor et al, 6; Massey et al, 6). This is ideal for studying receptor receptor interactions since a monomer has approximate dimensions of Å Å 7 Å. b ARs were labelled with stoichiometric amounts of either Cy or Cy5, and the efficiency of labelling was determined by absorption spectroscopy (Supplementary Table I). Labelled b AR was reconstituted into a mixture of,-dioleoyl-sn-glycero-- phosphocholine (DOPC) and cholesterol hemisuccinate (CHS) lipids. Three samples were generated for each experiment: () reconstituted Cy-labelled b AR; () reconstituted Cy5-labelled b AR and () Cy-labelled b AR mixed with Cy5-labelled b AR and then reconstituted. The final lipid-toreceptor molar ratio (mol-to-mol) was : unless otherwise indicated. The same samples were prepared for controls, but were maintained in.% DM and not reconstituted into vesicles. Orientation of b AR in lipid vesicles Knowing the orientation of b AR in our lipid vesicles is essential for interpreting FRET measurements. Random orientation would generate potential non-physiological (antiparallel) oligomers. While random orientation might be expected, previous studies have shown that rhodopsin orients predominantly in one direction following reconstitution (Niu et al, ). We used several complementary strategies to determine the orientation of b AR in phospholipid vesicles (Figure A). Factor Xa is a protease that selectively cleaves the b AR within the third ICL (ICL). Receptors oriented inside-out (ICL outside of lipid vesicle) will be susceptible to Factor Xa, whereas those oriented outside-out will not (Figure A). Approximately 9% of reconstituted b AR was TM6 (A65C) TM7 H8 (RC) TM5 TM TM TM TM ICL (T66C) Cy maleimide (donor) Cy5 maleimide (acceptor) Figure b AR single-cysteine constructs and FRET donor acceptor pair. (A) Three single-reactive cysteines constructs were generated on a minimal cysteine background (D5-b AR). The labelling sites were placed in the first ICL, D5-b AR-T66C, at the cytoplasmic end of the sixth transmembrane segment, D5-b AR-A65C, and helix eight, D5-b AR-RC. (B) Intracellular D view of the distribution of regions chosen for single-cysteine mutants, a-carbons are depicted. (C) FRET donor (l ex ¼ 59 nm; l em ¼ 57 nm) and acceptor pair (l ex ¼ 65 nm; l em ¼ 67 nm). & 9 European Molecular Biology Organization

3 b -Adrenoceptor oligomers PNGase F Outside of vesicle M antibody NHS-(PEO) - Biotin kda 97 Factor Xa βar Lipid vesicles + βar detergent + 5 α-flag Inside of vesicle Factor Xa C65-mBBr 9 βar Lipid vesicles βar Detergent PNGase F + + NHS-PEO Biotin 5 kda α-flag 5 kda 5 kda α-flag NHS-PEO Biotin Figure b ARs are predominantly oriented outside-out in lipid bilayers. (A) Strategies for determining orientation of b AR in lipid bilayers. (B) Purified receptors were reconstituted as described under Materials and methods and then subjected to treatment with Factor Xa and resolved by % SDS PAGE and transferred onto nitrocellulose. The presence of b AR was determined by probing with an M antibody conjugated with Alexa-68. (C) Samples subjected to PNGase F were prepared and imaged as in panel A. (D, E) Reconstituted samples were treated with the hydrophilic, amine-reactive, alkylating reagent NHS-PEO -biotin that disrupts binding of the M monoclonal antibody to the FLAG epitope. Samples were assessed for reactivity to M antibody (D) and an antibody that recognizes the C-terminal six-histidine tag (E). All data are representative of three independent experiments His resistant to Factor Xa, whereas all of the receptor was cleaved in the presence of.% DM (Figure B), a concentration of detergent, which permeabilizes the vesicles. These results are consistent with a predominantly outside-out orientation. PNGase F is an enzyme that cleaves asparagine-linked oligosaccharides on the extracellular N-terminus (Figure A). Treatment of reconstituted receptor with PNGase F led to a mobility shift that was indistinguishable from that observed in the presence of.% DM, consistent with predominantly outside-out orientation (Figure C). The orientation was further confirmed using NHS-PEO -biotin to chemically modify the N-terminus (Figure A). This polar compound would not be expected to cross the lipid bilayer. Chemical modification of the N-terminal FLAG epitope results in loss of reactivity to the M antibody; treatment of vesicles following reconstitution resulted in the loss of M reactivity for more than 9% of reconstituted b AR (Figure D and E). Finally, we labelled A65C on the cytoplasmic side of the b AR with monobromobimane (mbbr) and examined the ability of tryptophan in solution to quench bimane fluorescence. We observed no quenching of reconstituted, bimanelabelled b AR with mm tryptophan. However, solubilization of vesicles using.% DM resulted in significant quenching (Supplementary Figure ). Taken together, these studies show that the b AR is predominantly oriented with the extracellular domains on the outside of the vesicle. Distribution of b AR in lipid vesicles In studying oligomerization, it is important to avoid forcing protein together by inhomogeneous reconstitution, that is, trapping the majority of the receptor molecules in a minor population of lipid vesicles. For instance, it has been shown previously that 9% of rhodopsin molecules were incorporated into only % of vesicles (Mansoor et al, 6). We used isopycnic density centrifugation to assess the distribution of b ARs in lipid vesicles as previously described for rhodopsin (Mansoor et al, 6). Cy5-labelled b ARs were reconstituted at a lipid-to-receptor ratio of : in lipids containing NBD phosphocholine (at a final of.% of total lipid content). This allowed us to analyze samples subjected to a discontinuous sucrose density gradient by following Cy5 fluorescence (for the presence of b AR) and NBD fluorescence (for the presence of lipid vesicles). Our results show nearly perfect correlation between Cy5 fluorescence and NBD fluorescence at every fraction analyzed, suggesting that b AR molecules are uniformly distributed in these vesicles (Figure A). Similar results were obtained with b AR reconstituted at a : lipid-to-receptor ratio (Supplementary Figure ). To assess the density of b ARs in the lipid vesicles, we used electron microscopy to determine the average diameter of our b AR-containing lipid vesicle preparations. Using a negative staining protocol, we determined that the average diameter of our vesicle preparations at a lipid-to-receptor ratio of : was 8 nm± nm (Figure B). Using the calculations detailed in the Supplementary Materials and methods section, we concluded that there are 5 6 b ARs per lipid vesicle, with the majority oriented in an outside-out manner. Functional characterization of b AR in lipid vesicles We performed saturation binding on purified, reconstituted receptor to determine the affinity of all three single-cysteine mutants for the antagonist [ H]-dihydroalprenolol (DHA). We observed no significant difference between the three modified b ARs and wild-type b ARs (Table I and Supplementary Figure ). Competition binding studies with [ H]-DHA were used to determine the K i values for the agonist isoproterenol (Iso) and the inverse agonist ICI 8,55 (ICI). As shown in Table I and Figure, the values for the single-cysteine mutants are comparable to those obtained for wild-type & 9 European Molecular Biology Organization

4 b -Adrenoceptor oligomers Normalized F6 nm NBD-phosphocoline NBD Cy Normalized F 65 nm Cy5-β AR. 5 6 Fraction. % 6% % 8% % % Percent sucrose Figure b ARs are homogenously distributed in lipid vesicles. (A) To determine the distribution of b ARs in lipid vesicles, sucrose density gradients of samples containing.% NBD phosphocholine and Cy5 b ARs reconstituted at a lipid-to-receptor ratio of : were performed as described in the Supplementary data. Detection of lipid fractions was performed by following NBD fluorescence (l ex ¼ 6 nm) and receptor fractions by following Cy5 fluorescence (l ex ¼ 69 nm). (B) Reconstituted b ARs were imaged using a negative staining protocol as described in the Supplementary data to determine the size distribution of vesicles and the number of receptors per vesicle. Scale bar length represents nm. Data are representative of three independent experiments. Table I Agonist, antagonist and inverse agonist binding properties for the single-reactive cysteine receptors a b AR K i [s.e. interval] (nm) K d ±s.e.m. Mutant ( )-Isoproterenol ICI 8,55 [ H-DHA] Wild type 55 [ 6].7 [.77.77].±.6 D5-T66C 88 [9 7].9 [.86.5].8±. D5-A65C 98 [55 8].8 [.5.].5±. D5-RC [68 7].9 [.59.7].±. a Saturation and competition binding were performed as described under Materials and methods. Data represent the mean±s.e.m. of at least three independent experiments. b AR, suggesting that introduction of the single-reactive cysteines and reconstitution of purified b AR into lipid vesicles does not alter the pharmacology of the receptor. Functionality for G protein coupling of the three singlecysteine mutants was addressed by [ 5 S]-GTPgS binding. This assay involves reconstituting purified b AR with purified Tet-G as as previously described (Swaminath et al, 5; Granier et al, 7). Agonist binding to all three b AR single-cysteine mutants led to significant stimulation of G protein coupling that was similar to wild-type receptor (Figure C). Treatment of samples with the inverse agonist ICI led to decreases in basal activity similar to that observed for wild-type receptor (Figure C). Modification of the single cysteines with Cy5 maleimide fluorophore had no significant effect on G protein coupling (Figure C; P.5). FRET analysis of fluorophore-labelled b ARs in lipid bilayers We first determined FRET between receptors labelled at the same position. D5-A65C labelled with Cy was reconstituted with an equivalent amount of D5-A65C labelled with Cy5 in order to monitor TM6/TM6 interactions. This was repeated for D5-T66C and D5-RC, as reporters for ICL/ICL and H8/H8 interactions, respectively. Figure 5A shows an example of a typical experiment performed on D5-T66C. FRET between Cy- and Cy5-labelled receptors (.±.%) is only observed after receptor reconstitution into a lipid bilayer, but not when receptors remain solubilized in detergent (Figure 5A and Table II). Similar observations were made for Cy- and Cy5-labelled D5-A65C (6.7±.%) and for Cy- and Cy5-labelled D5-RC (6.9±.8%; Table II). To provide additional information about the relative orientation of b AR protomers, we investigated FRET between different labelling sites. For example, D5-T66C labelled with Cy was reconstituted with an equivalent amount of D5-A65C labelled with Cy5 in order to examine ICL/TM6 interactions. The same approach was followed for the other possible combinations, ICL/H8 and TM6/H8 (Table II). The observation of different FRET efficiencies for different labelling pairs suggests a specific arrangement of receptors in the lipid bilayers rather than nonspecific aggregation. To further rule out the possibility that the FRET observed in these studies is simply due to crowding of labelled receptors at the lipid bilayer, a -fold higher molar concentration of lipids (a final lipid-to-receptor ratio of :) was used in order to reduce the number of receptors per unit area of lipid bilayer. FRET efficiencies observed at a lipid-to-receptor ratio of : were not significantly different from those obtained at a ratio of : (Figure 5B D; P.5). FRET saturation of fluorophore-labelled b AR oligomers To further investigate the specificity of the observed oligomerization, as well as the stoichiometry of the oligomers, we performed FRET saturation experiments where the ratio of acceptor fluorophore (Cy5-labelled b AR) to donor fluorophore (Cy-labelled b AR) is increased, while maintaining the overall receptor concentration and lipid-to-receptor ratio constant. If the energy transfer is due to specific receptor receptor interactions, FRET efficiency will saturate as the Cy5/Cy ratio is increased. In contrast, random collisions should yield a quasi-linear relationship (Mercier et al, ; James et al, 6; Harikumar et al, 8). We observe FRET saturation for all three b AR labelling sites (Figure 6A C), demonstrating the specific nature of the interactions. In addition, FRET saturation can provide insight into the number of protomers per oligomer. Our FRET saturation results were compared with a well described mathematical model (Veatch and Stryer, 977; Mercier et al, ; James et al, 6; Harikumar et al, 8; Harding et al, 9) that & 9 European Molecular Biology Organization

5 b -Adrenoceptor oligomers % [ H]-DHA binding 8 6 WT 5-T66C 5-A65C 5-RC [Isoproterenol] (log M) % [ H]-DHA binding 8 6 WT 5-T66C 5-A65C 5-RC [ICI 8,55] (log M) 5 S-GTP γs bound (fold over basal) Isoproterenol + ICI 8,55 WT 5 A65C Cy5-5 A65C 5 T66C Cy5-5 T66C 5 RC Cy5-5 RC Figure Single-reactive cysteine mutants are fully functional. The affinity of the agonist isoproterenol (A) and the inverse agonist ICI 8,55 (B) was measured for all three single-cysteine mutants (D5-T66C, D5-A65C and D5-RC) and wild-type receptor by competitive binding of [ H]-DHA. Results are expressed as percent of radio-ligand bound in the absence of competitor. (C) Functionality of the three single-cysteine mutants, unlabelled or labelled with Cy5, and wild-type receptor was determined by GTPgS binding as described in the Supplementary data. [ 5 S]-GTPgS-specific binding induced by mm isoproterenol (agonist response) or by mm ICI 8,55 (inverse agonist response) is shown as fold over basal. All functional data represent the mean±s.e.m. of three independent experiments performed in triplicate. 5 Normalized fluorescence intensity 5 5 Before reconstitution (detergent micelles) After reconstitution (lipid bilayers) λ (nm) % FRET efficiency ICL/ICL : : % DDM % FRET efficiency TM6/TM6 : : % DDM % FRET efficiency H8/H8 : : % DDM Figure 5 Intermolecular FRET between Cy- and Cy5-labelled b AR is independent of other cellular proteins and is specific. (A) Purified, detergent-solubilized receptor protein was labelled with Cy or Cy5 maleimide and unreacted fluorophore was quenched with cysteine and separated from protein by gel filtration as described under Materials and methods. Cy- and Cy5-labelled protein samples were mixed at a : molar ratio and reconstituted into phospholipids bilayers or maintained in detergent. Subtraction of the proper controls and normalization of the raw traces is described in the Supplementary data. Labelled b ARs were reconstituted at a -fold higher lipid-to-receptor ratio ( :) and FRET efficiency was measured for ICL/ICL (B), TM6/TM6 (C) and H8/H8 (D) interactions. Data are representative of at least three independent experiments (A) or represent the mean±s.e.m. of at least three independent experiments (B D). & 9 European Molecular Biology Organization 5

6 b -Adrenoceptor oligomers Table II FRET efficiencies in the absence of ligand and upon binding of agonist, neutral antagonist or inverse agonist a b AR region No ligand±s.e.m. +Isoproterenol±s.e.m. P-value +Alprenolol±s.e.m. P-value +ICI 8,55±s.e.m. P-value ICL-/ICL-.6±..± ± ±..* TM-6/TM-6 6.7±. 7.7±..6.8± ±..5** H-8/H ± ±..6.± ±.8.8 ICL-/TM-6.58±..6± ±6..5.±.6.69 ICL-/H-8.8±.6.7± ± ±..6** TM-6/H-8.98±.7 5.7±.5.*.6±..5* 5.±..5** a FRET efficiencies were calculated as described under Materials and methods. Data represent the mean±s.e.m. of at least three independent experiments. P-values refer to statistical comparisons between no ligand and three different ligands: isoproterenol, alprenolol and ICI 8,55. *Po.5; **Po.5. % FRET efficiency ICL/ICL [Cy5]/[Cy] % FRET efficiency TM6/TM [Cy5]/[Cy] % FRET efficiency H8/H [Cy5]/[Cy]. Maximal FRET efficiency [Cy5]/[Cy] βar experimental data Dimer Trimer Tetramer Higher-order oligomer Figure 6 Specificity of b AR oligomerization as assessed by FRETsaturation. FRETsaturation involved varying the ratio of Cy5- to Cy-labelled b ARs over a range of : to : (Cy5:Cy), while the overall b AR concentration was kept constant. Saturable FRET is observed for ICL/ICL (A), TM6/TM6 (B) and H8/H8 (C). FRET measurements were performed and calculated as described in the Supplementary data. Data represent the mean±s.e.m. of at least three independent experiments. (D) FRET saturation data from all three constructs (A C above) was normalized to maximal FRET efficiency and then averaged and plotted together with theoretical curves (dashed lines) for dimer, trimer, tetramer and higherorder oligomer that were generated using equation () in the Supplementary data. has been used to predict the maximal energy transfer expected in energy transfer saturation experiments (FRET or BRET) for dimers, trimers, tetramers, etc. It follows that saturation will occur at a lower acceptor/donor ratio for higher-order oligomers than for simple dimers. We normalized our FRET saturation data for all three constructs and compared them with models for dimers, trimers, tetramers and higher-order oligomers (eight protomers), and found that our data are superimposed on the theoretical curve for tetramers (Figure 6D). The effect of ligand efficacy on b AR oligomerization We examined the effects of three classes of GPCR drugs: a full agonist (isoproterenol), a neutral antagonist (alprenolol) and an inverse agonist (ICI) on FRET efficiency between different labelling sites. Upon treatment with saturating amounts ( mm) of the full agonist isoproterenol, a small, but significant, increase in FRETwas observed between TM6 and H8 (Figure 7A and Table II; Po.5). At saturating concentrations (5 nm), alprenolol produced a similar result between TM6 and H8 (Figure 7A and Table II; Po.5). It is not possible to say if these small changes are due to subtle changes in the relative arrangement of protomers, or small conformational changes in the receptor. In contrast to the small changes observed with the agonist and neutral antagonist, much larger changes were observed following exposure to the inverse agonist ICI (Figure 7A and Table II). Inverse agonists include many compounds that were originally classified as antagonists, ligands that occupy the orthosteric binding site, but do not alter receptor function. Instead, inverse agonists inhibit basal agonist-independent activity exhibited by many GPCRs, including the b AR (Galandrin and Bouvier, 6). Interestingly, at a saturating concentration (5 nm) of ICI, significant changes in FRET efficiency were observed for four of the six labelling pairs (Figure 7A and Table II). The ICI-induced changes in FRET reach a maximum at min (Supplementary Figure ). The changes in FRET observed with ICI could reflect changes in the orientation of protomers or the number of protomers in the oligomeric complex. However, these changes could also be due to ligand-induced changes in the 6 & 9 European Molecular Biology Organization

7 b -Adrenoceptor oligomers % Change in FRET efficiency 6 * +ICI 8,55 +Isoproterenol +Alprenolol ** ** ICL ICL TM6 TM6 H8 H8 ICL TM6 ICL H8 * * ** TM6 H8 Maximal FRET efficiency..8 Higher-order oligomer.6 No ligand. +ICI 8,55 +Isoproterenol. +Alprenolol [Cy5]/[Cy] β AR labeling region Time (min) Bis (NHS)PEO Isoproterenol + + ICI 8, Oligomer Monomer Cy5 fluorescence kda Maximal FRET efficiency Higher-order oligomer No ligand +Carvedilol +Carazolol [Cy5]/[Cy] Figure 7 b AR oligomers are regulated by inverse agonists. (A) Treatment of FRET samples with saturating amounts of the inverse agonist ICI 8,55, agonist isoproterenol and neutral antagonist alprenolol. (B) FRET saturation in the presence of ligands. Isoproterenol and alprenolol led to no observable difference from the unliganded FRET saturation curve, whereas ICI 8,55 yielded to a curve that is more consistent with higher-order oligomers. (C) Cross-linking of reconstituted Cy5-labelled b AR samples in the presence or absence of isoproterenol or ICI 8,55 was carried out as described in the Supplementary data. (D) FRET saturation in the presence of the inverse agonists carvedilol (red) and carazolol (green). All data are reported as mean±s.e.m. (A, B, D) or are representative of at least three independent experiments (C). *(Po.5) and ** (Po.5). conformation of receptors that influence the mobility of the fluorophore or the polarity of its environment. We, therefore, examined the effect of isoproterenol and ICI on the intensity of the fluorescence and on the anisotropy of the fluorophores in labelled receptors. For both Cy- and Cy5-labelled b AR reconstituted individually, treatment with either isoproterenol or ICI did not induce significant changes in the intensity of the fluorescence (data not shown) or anisotropy of the fluorophores, suggesting that the change in FRET efficiencies observed upon ICI treatment are not a result of conformational changes in protomers (Supplementary Figure 5). The ICI-induced changes in FRET efficiency may be attributed to several additional factors: reorientation of protomers in the oligomers, a tighter packing of the protomers, an increase in the temporal stability of the oligomers (assuming there is an equilibrium between monomers and oligomers), and an increase in the stoichiometry of the oligomeric state (e.g., going from dimers or tetramers to higher-order oligomers). To distinguish between these possibilities, we performed FRET saturation in the presence and absence of ICI, alprenolol or isoproterenol. Samples were incubated with ligands for min at room temperature and measurements were taken. Results show that in the presence of ICI the saturation curve is more similar to a model for higher-order oligomers, while alprenolol and isoproterenol appear to have no effect on the apparent oligomeric state of the receptor (Figure 7B). Higher-order oligomerization was also observed for the inverse agonists carazolol and carvedilol (Figure 7D). Cross-linking was used to further address the state of multimeric assembly of the b AR. We used Bis(NHS)PEO 5, a homobifunctional cross-linker with a spacer length of.7 Å that covalently modifies e-amines of lysine residues and a-amine groups at the N-termini, effectively trapping receptors that come within interacting distances. Although concerns have been raised about the potential for crosslinkers to trap transiently interacting proteins (Brett and Findlay, 979; Downer, 985; Medina et al, ), it is evident that pre-incubation of samples with ICI leads to more extensive cross-linking and trapping of higher-order oligomers of reconstituted b AR when compared with the unliganded, the agonist and the antagonist treated samples (Figure 7C and Supplementary Figure 6). Taken together, these results suggest that the b AR forms higher-order oligomers in the presence of the inverse agonists ICI, carazolol and carvedilol. To investigate the effect of ICI on the stability of interactions between protomers, we monitored FRET following addition of.% DDM, a concentration of detergent that solubilizes the vesicles. We found that the decline in FRET following the addition of detergent was slower and less complete in samples pre-incubated with ICI compared with unliganded samples (Supplementary Figure 7, Po.5), providing evidence that ICI also stabilizes interactions between protomers. & 9 European Molecular Biology Organization 7

8 b -Adrenoceptor oligomers The effect of Gs on b AR oligomerization To investigate the effect of G protein coupling on oligomerization, we performed FRET saturation experiments by reconstituting b AR with a three-fold molar excess of purified Gs heterotrimer (Figure 8). This concentration of G protein was chosen to ensure that sufficient G protein would be incorporated into vesicles while having a minimal effect on the reconstitution. The inclusion of Gs did not alter the orientation of the b AR as determined by the susceptibility to PNGase F (Figure 8A). We observed a statistically significant (Po.8) decrease in FRET saturation in the presence of Gs as compared with reconstitutions in the absence of Gs (Figure 8B) for Cy5/Cy of.5,.5 and. To determine whether the effect of Gs on FRET saturation was due to A B C D Maximal FRET efficiency Maximal FRET efficiency Normalized fluorescence intensity Vesicles + Gs Vesicles + % DDM PNGase F kda Cy5 fluorescence. [Cy5]/[Cy] Higher-order oligomer β AR + Gs + GTPgS Tetramer β AR + Gs Trimer. [Cy5]/[Cy] Higher-order oligomer β AR alone Tetramer β AR + Gs Trimer mbbr-β AR + Gs-GTPγS mbbr-β AR + Gs λ (nm) nonspecific effects of reconstituting with another membrane-associated protein, we performed FRET saturation of b AR in the presence of Gs and GTPgS, which uncouples b AR and Gs. As shown in Figure 8C, the presence of GTPgS increases, in a statistically significant manner (Po.), FRET saturation to the values observed for b AR alone. To estimate the fraction of b AR that couples to Gs under these reconstitution conditions, we labelled C65 at the cytoplasmic end of TM6 with mbbr, an environmentally sensitive fluorescent probe. We previously showed that maximal coupling of Gs to b AR reconstituted into HDL particles results in a decrease in the fluorescence intensity and an 8-nm shift in the maximal emission wavelength (l MAX ) of mbbr b AR (mbbr b AR; Yao et al, 9). As shown in Figure 8D, under reconstitution conditions used for FRET saturation experiments, Gs induced a decrease in intensity and a -nm shift in l MAX of mbbr b AR relative to the same reconstitution in the presence of GTPgS. Based on the shift of l MAX we estimate that approximately % of the reconstituted b AR is coupled to Gs. Discussion Receptor dimerization plays an essential role in the function of Family C GPCRs (Margeta-Mitrovic et al, ; Pin et al, 5). However, the role of oligomerization for Family A (rhodopsin-like) GPCRs is less clear. It has been shown that monomeric rhodopsin and b AR can activate their respective Gproteins(Bayburtet al, 7; Ernst et al, 7; Whorton et al, 7). Yet, there is convincing evidence from a variety of experimental approaches that the b AR and many other Family A GPCRs exist as dimers or oligomers in the plasma membrane. Most compelling are studies that apply FRETand/ or BRET technology to receptors tagged with fluorescent proteins and expressed in cultured cells (Angers et al, ; Mercier et al, ; Milligan and Bouvier, 5; Guo et al, 8), as well as studies using disulphide cross-linking to trap interactions in cell membranes to map the interface between protomers (Guo et al,, 5, 8; Klco et al, ). Characterization of the structure (protomer organization) and dynamics of GPCR oligomers is challenging and will Figure 8 Effect of the G protein Gs on FRET saturation of Cy5- and Cy-labelled b AR. FRET saturation was performed by varying the ratio of Cy5- to Cy-labelled b AR-RC over a range of : to : (Cy5:Cy), while the overall b AR concentration was kept constant. Purified Gs heterotrimer was added at a molar ratio of Gs: b AR before reconstitution. (A) The inclusion of Gs in the reconstitution did not alter the orientation of b AR in vesicles as determined by the susceptibility of reconstituted b AR to PNGase F (see Figure C). FRET saturation was significantly lower in the presence of Gs compared with b AR alone (B) or b AR and Gs with mm GTPgS (C). (D) b AR was labelled on C65 at the cytoplasmic end of TM6 with mbbr b AR and reconstituted with Gs under the same conditions that were used for FRET saturation experiments. Gs induced a decrease in intensity and a -nm shift in l MAX of mbbr b AR relative to the same reconstitution in the presence of GTPgS. A two-way ANOVA was used to compare FRET values for b AR, b AR þ Gs and b AR þ Gs þ GTPgS at the different Cy5:Cy ratios. A posteriori statistical analysis showed significant decrease in FRET between b AR and b AR þ Gs (Po.8), and a significant increase in FRET between b AR þ Gs and b AR þ Gs þ GTPgS (Po.) for all Cy5:Cy ratios except and. No statistical differences are found between b AR and b AR þ Gs þ GTPgS. 8 & 9 European Molecular Biology Organization

9 b -Adrenoceptor oligomers require integration of information from a variety of different approaches. In an effort to provide additional structural insight into the organization, stability and regulation of GPCR oligomers by ligands, we used FRET to study oligomerization of purified b AR site specifically labelled with relatively small fluorescent probes and reconstituted into a model lipid bilayer. Our results show that monomeric b ARs oligomerize spontaneously upon reconstitution into lipid bilayers in the absence of other cellular chaperones or scaffold proteins. FRET saturation studies suggest that the oligomers consist of more than two protomers, and are probably tetramers. Agonists have little effect on b AR oligomerization, whereas inverse agonists appear to promote higher-order oligomerization and stabilize the oligomers against dissolution by detergent. Spontaneous oligomerization of the b AR in model lipid membranes In a cell membrane, proteins associate in a complex lipid environment involving mixtures of cholesterol and lipids having different polar groups, alkyl chain lengths and alkyl chain saturation. Moreover, the lipid composition of outer and inner membrane layers is different, and there may be distinct lipid domains that regulate the function of associated membrane proteins (Allen et al, 7). b AR oligomers have been observed to form during biosynthesis in the endoplasmic reticulum (Salahpour et al, ), suggesting that chaperones and other cellular proteins may also be involved in the assembly and/or maintenance of oligomers for some GPCRs. For example, oligomerization of the opioid receptor has been shown to depend on the G protein Gi (Law et al, 5). Nevertheless, our results show the strong tendency for purified b AR to oligomerize spontaneously in a model lipid bilayer, and suggest that b AR oligomerization is an intrinsic property of the receptor and does not require other cellular proteins or a specific lipid environment. However, we cannot exclude the possibility that oligomerization in vivo is regulated in some way by cellular proteins or specific lipid environments. It should be noted that rhodopsin, neurotensin receptor and muscarinic receptors have also been observed to oligomerize following reconstitution into synthetic lipid bilayers (Mansoor et al, 6; Ma et al, 7; Harding et al, 9), suggesting that spontaneous oligomerization might be an intrinsic property of Family A GPCRs. Following purification from insect cell membranes, b ARs exist as pure monomers in detergent solution (Whorton et al, 7). Our reconstitution experiments use a simple model lipid bilayer composed of DOPC and CHS (DOPC/CHS). In this environment we observe normal ligand binding properties and efficient G protein activation. The high degree of homogenous orientation of receptors in this lipid bilayer model suggests that there must be a non-random mechanism that controls receptor insertion. While the mechanism for this is unknown, preferential orientation has also been observed with reconstitution of rhodopsin, although in this case receptor was oriented preferentially with the N-terminus in the inside of the vesicle (Niu et al, ). Isopycnic density centrifugation experiments and EM images (Figure ) show that reconstitution using size-exclusion chromatography yields a homogenous distribution of receptors into lipid vesicles of an average diameter of 8 nm and an average surface area of approximately 6 nm. Under our reconstitution conditions, we estimate that there are 5 6 b AR molecules per vesicle. Based on the crystal structure of the b AR, we can calculate that the surface area occupied by a single receptor is approximately 6 nm. Therefore, receptors occupy less than % of the surface area of the vesicle, suggesting that the FRET we observe is not the result of nonspecific interactions due to high receptor density. To further exclude the possibility of nonspecific oligomerization, we reconstituted receptors at a -fold higher lipid-to-protein ratio and found no significant decrease in FRET efficiency (Figure 5B D, P.5). Organization and stability of b AR oligomers Our reconstitution system, as well as current cell-based methods, is limited in the ability to precisely define the structure and stoichiometry of oligomers. Nevertheless, recent evidence from resonance energy transfer studies, cysteine cross-linking and fluorescence recovery after photobleaching (FRAP) studies suggest that the D dopamine receptor and the b AR form higher-order oligomers (tetramers or greater) (Guo et al, 8; Dorsch et al, 9). Data from FRET studies can provide insight into the orientation of protomers within these oligomers. However, the following facts should be taken into account. First, while Cy and Cy5 are small relative to fluorescent and luminescent proteins, they are still large compared with amino acids and have relatively long flexible linkers tethering them to cysteine (Figure 9A). As such, their precise orientation in the protein is not known and has been estimated by computational techniques. Second, both the distance between fluorophores and the orientation of fluorophores influence FRET. The orientation factor becomes a concern for highly constrained fluorophores; however, this probably is not the case here, since the anisotropy values that we observe experimentally for Cy and Cy5 are similar to those reported for free fluorophores in solution (Kobitski et al, 7), and these values are not influenced by ligands. Third, as discussed below, the oligomers may be dynamic. While the predominant form may be a tetramer, these may exist in equilibrium with monomers, dimers and higher-order oligomers. Finally, in a tetramer several combinations of donor/acceptor pairs ( donors, acceptors; donor, acceptors; etcy) are possible. In such a system, the measured FRET signal will be the combination of all the individual energy transfers between every possible donor/acceptor pair, which complicates the analysis of the measured FRET signal, and thus, the estimation of protomer orientation in the tetramer. Notably, saturation experiments (Figure 6) result in a much simpler system, composed mostly of tetramers with one donor and three acceptors. This system dramatically reduces the number of energy transfer combinations and simplifies the analysis of FRET efficiencies. These values are used to propose possible monomer orientations within the tetramer. In FRET saturations studies, the greatest energy transfer is observed for H8/H8 and smallest for TM6/TM6 (Figure 6B and C). These results are compatible with arrangements of oligomers involving a TM interface (Figure 9C) previously described for the D dopamine receptor and rhodopsin (Liang et al, ; Guo et al, 8). In contrast, an arrangement that would place TM6/TM6 fluorophores in close proximity (Figure 9B) is not compatible with the lowest FRET efficiency observed for this pair (Figure 6B). & 9 European Molecular Biology Organization 9

10 6 b -Adrenoceptor oligomers Inactive H8 Active H H8 ICI H H8 ISO ICI 5 7 H8 5 7 H H H H H H8 Figure 9 Schematic representation of possible b AR oligomers. (A) Cytoplasmic view of the D structure of the b AR (left) and cartoon of this footprint (right), with the centre of mass of Cy depicted as spheres: T66C (green), A65C (blue) and RC (red). TM6 is depicted in the inactive conformation (top) and in the proposed active conformation as observed in the structure of opsin (bottom). (B) Receptor oligomerization involving the surfaces of TMs or/and 5 is not compatible with our FRET results and might sterically prevent movement of TM6. (C) Our FRET results suggest an arrangement of protomers involving the TM interface. (D) The movement of the cytoplasmic end of TM6 upon agonist binding repositions the fluorophore outwards in two of the four protomers, and towards H8 in the other two. (E) Treatment with the inverse agonist ICI probably reduces conformational fluctuations responsible for basal activity, increasing the packing of the oligomers. (F) ICI may stabilize higher-order oligomers where TM6 is packed into the core of the oligomer, contributing to increase FRET efficiency and possibly the inactive state of the receptor. While our FRET saturation experiments best fit a mathematical model for tetramers (Figure 6D), our results are compatible with a dynamic equilibrium where a fraction of b AR exists as monomers in equilibrium with higher-order oligomers, with the average size of the oligomer being a tetramer. This is in agreement with the observed maximal FRET saturation values of 5% (Figure 6), lower than would be expected for stable tetramers given that R values for Cy and Cy5 range from 7 56 Å, depending on the biochemical system under investigation (Mansoor et al, 6; Massey et al, 6). Further evidence supporting this dynamic behaviour is the observation that the affinity of interactions between protomers is relatively weak outside of the bilayer environment, as show by the rapid dissociation of protomers upon the addition of a non-ionic detergent (Supplementary Figure 7). Recent FRAP studies provide evidence that D dopamine receptors may also exist in a dynamic equilibrium of monomers and oligomers (Fonseca and Lambert, 9). In these studies, oligomers were only detected by FRAP after receptors formed stable covalent dimers through oxidative cross-linking of cysteines in TM. While a b AR monomer can activate Gs (Whorton et al, 7), it is not known whether higher-order oligomers facilitate or impair coupling. Oligomers of rhodopsin (Bayburt et al, 7)andNTreceptor(Whiteet al, 7) couple less efficiently to G proteins than monomers. If higher-order oligomers impair coupling, the dynamic character of b AR oligomers would ensure that a fraction of the b AR would exist as monomers or dimers competent for G protein activation. Under our experimental conditions, the co-reconstitution of Gs with b AR was associated with a small decrease in FRET saturation that was reversed by GTPgS (Figure 8). This is compatible with G protein coupling shifting the equilibrium to lower-order oligomers. & 9 European Molecular Biology Organization

11 b -Adrenoceptor oligomers Ligand regulation of b AR oligomers The effect of agonists on oligomerization has been described for several GPCRs. The results appear to be receptor specific: for some receptors no effect is observed, whereas for others, agonists induce dissociation or association (Angers et al, ; Cheng and Miller, ; Latif et al, ; Zhu et al, ; Roess and Smith, ; Dorsch et al, 9). For the b AR, Michelle Bouvier s laboratory reported a small agonistinduced increase in steady-state BRET; however, they concluded that this could be due to a small change in the steadystate oligomers or due to conformational changes in individual protomers (Angers and Bouvier, ). In subsequent BRET saturation studies from this laboratory (Mercier et al, ) and fluorescence recovery after photobleaching studies from the Bünemann laboratory (Dorsch et al, 9), no significant agonist-induced effect was observed. Our results with the full agonist (isoproterenol) are in agreement with these cell-based studies. Isoproterenol causes a relatively minor change in intermolecular FRET, with the only significant change occurring in the TM6/H8 FRET pair (Figure 7A and Table II), but no change in FRET saturation (Figure 7B). The results are in agreement with the model depicted in Figure 9C and D where agonists induce a change in protomer conformation, but not oligomerization. We have previously shown significant agonist-induced movement of TM6 (Gether et al, 997; Ghanouni et al, a, b; Yao et al, 6) consistent with relatively large conformational changes observed in rhodopsin by DEER spectroscopy (Altenbach et al, 8), and from the crystal structure of opsin (Park et al, 8; Scheerer et al, 8). Assuming a tetramer, the movement of the cytoplasmic end of TM6 away from TM (Park et al, 8) (Figure 9A) locates the fluorophore attached to TM6 towards the periphery in two of the four protomers, and towards H8 in the other two (Figure 9D). Thus, the symmetric movement of TM6 in the tetramer, inwards and outwards, explains the minor changes in intermolecular FRET of TM6/TM6 observed upon agonist binding. In addition, the inward movement of TM6 towards H8 is compatible with the changes occurring in the TM6/H8 FRET pairs. Therefore, protomer packing does not appear to interfere with conformational changes involving TM6. In agreement with this, agonist-induced movement of the cytoplasmic end of TM6, as detected by an environmentally sensitive fluorophore covalently bound to C65, is similar for b AR monomers (reconstituted into HDL particles) and oligomers (reconstituted into phospholipids vesicles) (Supplementary Figure 8). We were surprised to see that the most dramatic changes in intermolecular FRET were observed on exposure to the inverse agonist ICI. Most notable are increases in FRET for TM6/TM6 and ICL/H8 (Table II and Figure 7A). Fluorescence studies on monomeric b AR may provide us with the link between the effect of an inverse agonist on the structure of the monomer and the process of oligomerization. We have previously shown that ICI does not induce major rearrangements in the structure of the monomer, but may reduce normal conformational fluctuations responsible for basal activity (Yao et al, 6, 9). We speculate that this inherent flexibility may interfere with higher-order packing of the oligomers. Thus, the more constrained structure of the inverse agonist-bound receptor may be more compatible with closer packing of protomers within a tetramer (Figure 9E), a higher-order packing (Figure 9F) or more stable b AR oligomers with fewer monomers. It has been observed that oligomers of rhodopsin (Bayburt et al, 7) and NT receptor (White et al, 7) couple less efficiently to G proteins than monomers; therefore, higher-order packing induced by the inverse agonist may restrict access of receptor to G protein. However, this higher-order packing is not required for the inverse agonist effect, as an inverse agonist can efficiently prevent coupling of monomeric b AR to Gs (Yao et al, 9). It is likely that a combination of all three effects is responsible for the ICI-induced FRET changes. The increased stability of oligomers against dissociation by detergent in the presence of ICI (Supplementary Figure 7) is in agreement with closer packing of protomers, while FRET saturation (Figure 7B and Supplementary Figure 5) and cross-linking experiments (Figure 7C and Supplementary Figure 6) are in agreement with higher-order oligomers. Our results are consistent with a rearrangement of the oligomerization interfaces that has been observed for the dopamine D receptor upon inverse agonist binding (Guo et al, 5), and the observation of higher-order packing of the inactive state of rhodopsin in rod outer segment membranes (Liang et al, ). The functional consequence of this higher-order oligomerization is not known, but could impair coupling of the b AR to Gs, or be involved in the coupling of the b AR to other signalling pathways. Interestingly, both ICI and carvedilol have been shown to activate MAPK through a G protein-independent, arrestin-dependent signalling pathway (Galandrin and Bouvier, 6; Wisler et al, 7). The dramatic effects of ICI on oligomerization might be expected to influence ICI binding affinity or cooperativity; however, no differences in ICI binding properties were observed between b AR monomers (reconstituted into HDL particles) and oligomers (reconstituted into phospholipids vesicles) (data not shown). In conclusion, we find that the b AR is capable of forming specifically oriented multimeric assemblies in a model lipid bilayer in the absence of other cellular proteins, complementing previous studies of Family A GPCRs. Although we cannot determine the oligomeric interfaces with precision, our results are compatible with models proposed for several other Family A GPCRs, where oligomerization involves primarily the TM/H8 interface. Most unexpected was the observation that inverse agonists promoted higher-order b AR oligomerization. This may alter access to other signalling proteins, providing insight into the ability of inverse agonists to inhibit basal G protein signalling or in promoting G protein-independent activation of MAPK pathways. These results suggest a potential structural link between the stabilizing effects of inverse agonists on b AR monomers and the assembly of oligomers in lipid bilayers. Materials and methods Materials Cy maleimide and Cy5 maleimide were both obtained from Amersham Biosciences. [ H]-DHA and [ 5 S]-GTPgS were purchased from Amersham and Perkin Elmer, respectively. All drugs tested were purchased from Sigma. The Bis(NHS)PEO 5 homobifunctional cross-linker and NHS-PEO -Biotin were purchased from Pierce. Engineering of single-cysteine b AR mutants Site-directed mutagenesis of the b AR was performed by using the human b AR cdna containing a FLAG epitope at the N-terminus as & 9 European Molecular Biology Organization